Controversy Extenders | X-Ray Transparency | Incidental PE AI

“Radiologists who train NPP to interpret imaging studies are arguably training their replacements.”

A JACR paper with a warning for radiologists about their non-physician colleagues.


I’m excited to welcome Arterys as an Imaging Wire sponsor. Arterys is doing amazing work weaving leading AI clinical applications directly into the medical imaging workflow, and I can’t wait to share their message with you.

Every issue of The Imaging Wire is made possible by Arterys and the rest of our sponsors, so make sure to keep up with the great things that they’re doing.

Imaging Wire Sponsors

  • Arterys – Transforming medical imaging by reducing the tedium of interpretation and empowering physicians with our web-based AI platform.
  • Bayer Radiology – Providing a portfolio of radiology products, solutions, and services that enable radiologists to get the clear answers they need.
  • GE Healthcare – Enabling clinicians to make faster, more informed decisions through intelligent devices, data analytics, applications and services.
  • Healthcare Administrative Partners – Empowering radiology groups through expert revenue cycle management, clinical analytics, practice support, and specialized coding.
  • Hitachi Healthcare Americas – Delivering best in class medical imaging technologies and value-based reporting.
  • Nuance – AI and cloud-powered technology solutions to help radiologists stay focused, move quickly, and work smarter.
  • Riverain Technologies – Offering artificial intelligence tools dedicated to the early, efficient detection of lung disease.
  • Siemens Healthineers – Shaping the digital transformation of imaging to improve patient care.

The Imaging Wire



RE Controversy

Last week’s UPenn study highlighting radiology extenders’ (REs) ability to speed up CXR workflows created quite a controversy, as many in the radiology community objected to what they viewed as an “REs can replace radiology residents” angle. Here are some details on the JACR study (that has since been erased from the internet) and the subsequent fallout.

  • The Study – The researchers had REs and residents each draft CXR reports for attending thoracic radiologists, finding that the REs significantly improved the radiologists’ productivity compared to when the radiologists produced the reports on their own (93.6 cases per hour vs. 62.4/hr). The radiologists’ productivity also improved with the resident-drafted reports, but not by as much (72.6/hr vs. 62.4/hr).
  • The Fallout – This was a quality assurance paper intended to evaluate REs’ efficiency in CXR reporting workflows, not compare their performance against residents (the sample was too small for that, anyway). However, that’s not how everyone took it, and the “REs can replace radiology residents” angle suddenly became the study’s unofficial conclusion. Folks on Twitter got mad, corporatization conspiracy theories started flying around the Aunt Minnie Forums, and the JACR pulled the study and quickly released a pro-radiologist paper (see the story below). Even the Imaging Wire got a request to remove our coverage (and I did acquiesce and make some edits).
  • The Takeaway – Most would agree that the way you prove the efficiency of something or someone (e.g. REs) at a task that has historically been performed by someone else (e.g. radiologists or residents) is by comparing the two while performing the same task. We see studies comparing AI against radiologists this way every single week. Still, this study unearthed some very real concerns among radiologists (who’ve seen how NPPs affected other specialties), and at least for now, the study is off the internet and people on Twitter are still upset.



RE Paradox

A new JACR paper warned radiologists that the growing role of non-physician practitioners (NPPs, radiology extenders, REs) could undermine their profession, while suggesting that the best way to keep NPPs from taking over more radiologist territory is to embrace AI. This requires some explanation, so here it is:

  • Background – The use of physician extenders has increased across many specialties, and radiology could be next as imaging volumes grow and reimbursements decline, potentially taking some work from radiologists. It’s also worth noting that this paper was scheduled for release a few months from now but after all the backlash after last week’s UPenn study on RE efficiency (see the story above) the JACR decided to fast-track its release.
  • Unintended Consequences – Although radiology departments and practices might be attracted to REs’ cost and efficiency advantages, the authors warn that expanding non-physicians’ role “could diminish physicians as healthcare providers altogether.”
  • Legislative Momentum – There’s been significant legislative progress in favor of NPPs operating independently (25 states now give NPPs full practice authority). Radiology hasn’t felt the effects of this movement so far (it’s mainly in anesthesiology and emergency medicine), but the authors warn that this could change if REs become payable providers and are allowed to operate independently.
  • Radiology, it’s Complicated – The authors also warned that the belief that using NPPs/REs for “simple tasks such as nasogastric tube placement” doesn’t take into account that radiologists are responsible for the whole image (not just identifying tips of tube or lines) and using an RE doesn’t remove the “liability for finding the unexpected.”
  • Cost Questions – Using REs for low-RVU / high-labor tasks might make sense for private practice radiologists due to profit sharing, but the paper warns that it might not help radiologists working under other employment models (academic, hospital, corporate). Because of that, the authors suggest that these larger entities might lobby for more NPP rights so they can reduce their radiologist headcounts.
  • The AI Solution – Noting the emergence of Medicare reimbursements for AI-based diagnoses (specifically, Viz.ai and RapidAI’s LVO tools), the authors suggest that AI could “easily be integrated into existing workflow(s)” and help radiologists gain RE-level efficiencies, still get paid for it, and rest easy knowing that legislation in favor of autonomous unsupervised AI isn’t coming soon.
  • AI Firsts – I’ve seen a lot of warnings that AI might replace radiologist jobs, but this is the first time I’ve seen a suggestion that AI would save radiologist jobs by replacing REs. This is also the first time I’ve ever seen a radiologist say that AI could be “easily be integrated into existing workflow.” Either way, AI companies have to be pumped about both of these statements.
  • The Takeaway – It’s become quite clear over the last week that rads vs. REs is a sensitive topic for everyone involved. It doesn’t appear that radiologists, REs, and institutions are ready to settle this, but at least we’re now talking about this issue in the open.



X-Ray Transparency

The humble X-ray took center stage in the ongoing federal healthcare cost transparency case, as appellate judges used the imaging exam as an example of how hospitals could be more transparent with their pricing than AHA lawyers claim is possible.

  • Unknowable X-ray – After an AHA litigator suggested that the rates hospitals negotiate with insurers are “unknowable,” the judges zeroed-in on X-rays. The judges detailed hypothetical scenarios where they received X-rays and paid a certain rate, asking the lawyers “why is the price of the X-ray unknowable?” The AHA explained that X-rays are billed in a variety of ways, making itemization difficult, but the judges weren’t buying it.
  • January’s Coming – The X-ray questioning made for some fun quotes, but the more important takeaway is that the hospital industry is relying on this case (its second appeal) to avoid mandatory healthcare cost transparency changes on January 1st. Based on Healthcare Dive’s observations, it’s “hard to see AHA prevailing.”

The Wire

  • AI for Incidental PE: Aidoc announced the FDA approval of its AI algorithm intended to flag and communicate incidental pulmonary embolism findings in lung CT scans, which is an industry-first, and comes about 5 months after the company’s incidental COVID-19 algorithm gained temporary FDA approval during the COVID-19 emergency. These incidental PE findings are reportedly most common during outpatient oncology imaging exams, allowing them to identify and diagnose PE while the patient is still in the building.
  • CTC Down at 65: Screening CT colonography (CTC) utilization rates increase at each age between 52 and 64 (5.3% per year), but fall at an even faster pace from 65 to 69 (−6.9% per year). That’s from a new JACR study (n = 12,648 CTC exams) that attributed the drop-off to the differences between commercial coverage (covers CTC) and Medicare (doesn’t cover CTC), suggesting that this decline has the greatest impact on racial minorities.
  • Serena Bright’s 510(k): GE Healthcare announced the FDA 510(k) approval and launch of its Serena Bright system, making it the first commercially available contrast-enhanced mammography (CESM) biopsy solution. Serena Bright allows clinicians to perform breast biopsy exams with the same mammography equipment, in the same room, and with the same staff as a screening or diagnostic mammogram exam.
  • Bracco’s New CardioGen-82: Bracco Diagnostics announced the FDA approval of its new CardioGen-82 Infusion System, which generates Rubidium Rb 82 for cardiac PET myocardial perfusion imaging. The successor to Bracco’s previous CardioGen-82 system (yes, same name) launches with improvements that focus on usability, efficiency, and radiation reduction.
  • Lesion-Aware: A Chinese research team announced the development of their Lesion-Aware CNN, capable of classifying 14 thoracic pathologies in chest x-rays and localizing regions where each pathology is suspected. Using a lesion-detection network (1,937 CXRs, 598 patients) combined with a lesion-classification network (8,801 CXRs, 2,928 patients), the Lesion-Aware CNN surpassed radiologists when identifying atelectasis (0.831 AUC vs. 0.807–0.855), mass (0.959 AUC vs. 0.944–0.974), and nodule (0.928 AUC vs. 0.906–0.950), while achieving statistically similar performance classifying the other 11 pathologies. The CNN also completed each CXR classification far faster than the radiologists (∼0.197 seconds vs. ∼35 seconds).
  • Ezra’s AI: Direct-to-consumer MRI cancer screening company, Ezra, announced the FDA approval of its new Prostate AI system. Prostate AI is designed to streamline Ezra’s partner radiologists’ prostate MRI analysis with a number of automated measurement and segmentation tools. This is a big milestone for Ezra, which has offered DTC prostate cancer screening since early 2019, and has maintained that AI efficiencies will be a key part of its business model for even longer than that.
  • POC MIR for ICH: Low-field point-of-care MRI systems (specifically Hyperfine’s 0.64T MRI) could be used to characterize intracerebral hemorrhage (ICH), potentially as an alternative to traditional MRI or CT systems. That’s from a new study presented at the American Neurological Association’s annual meeting that reviewed scans from 15 patients with ICH (low-field MRI, standard MRI, CT), finding that the portable MRI’s measurements were accurate compared to measurements from the standard MRI and CT systems (although more research is needed).
  • Radiologists’ Screening Influence: Patients with higher lung cancer risks are more likely to begin LDCT lung cancer screening after a shared decision making (SDM) session with a radiologist (93%) than with any other clinician (NPs = 80.1%, pulmonary specialists = 56.4%, family physicians = 53.9%). That’s from a new JAMA study of 11,699 Medicare enrollees who attended a SDM meeting, including 7,522 who went on to participate in LDCT screenings within three months. The study also found that patients were far less likely to begin LDCT screening if the patient and SDM clinician worked together in the past (55% vs. 75%).
  • Quantib’s Prostate AI 510(k): Quantib announced the FDA clearance of its prostate MRI solution, representing the Dutch developer’s 6th FDA clearance. Noting the massive growth in prostate MRI scans, the Quantib prostate solution comes with a suite of tools to streamline reporting (volumetry, PSA density, segmentation, and more).
  • Spectral CT’s Early COVID Advantage: A new AJR retrospective case review (four patients, two CT scans each) revealed that spectral CT (aka Dual-energy CT) might improve assessments of patients with early-stage COVID-19 compared to conventional CT. Spectral CT owes its early-COVID advantage to its superior visualization of ground-glass opacities (a weakness for conventional CTs), while it might also improve assessment of lung lesions.
  • RadNet & Adventist’s Simi JV: RadNet and Adventist Health agreed to create a joint venture that will initially include three outpatient facilities in Simi Valley, California (two existing RadNet centers, one Adventist center) and place Adventist Health’s Nancy Reagan Breast Center under RadNet’s management.

The Resource Wire

– This is sponsored content.

  • The Arterys Marketplace now includes 34 AI solutions available to its healthcare partners, including new FDA-cleared and CE-marked algorithms.
  • This Hitachi post details how human centered design led to the Scenaria View CT’s innovative Lateral Shift Table, which increases spatial resolution, allows accurate centering, and reduces dosage.
  • Radiologists have a new tool available that will potentially allow them to recover some of the added costs related to safety precautions taken during the COVID-19 PHE. Learn more about CPT code 99072 in this blog article from Healthcare Administrative Partners.
  • Check out how Siemens Healthineers’ Tin Filtration is like selecting better quality sunglasses.
  • In its latest Q&A, Nuance sat down with Infervision North America Director, Matt Deng, Ph.D., to discuss the company’s journey and the impact of its InferRead Lung CT solution.
  • This GE Healthcare case study details how Shields Health Care Group reduced its MRI scan times by 30% after adopting GE’s AIR Recon DL image reconstruction technology.
  • This Riverain Technologies case study details how Duke University Medical Center integrated ClearRead CT into its chest CT workflows, reducing read times by 26% and improving nodule detection by 29%.


Leave a Reply

Basic HTML is allowed. Your email address will not be published.

Subscribe to this comment feed via RSS

Get every issue of The Imaging Wire, delivered right to your inbox.

Join thousands of imaging professionals.